Authoring LLM-Based Assistance for Real-World Contexts and Tasks

I'm happy to announce that our paper Authoring LLM-Based Assistance for Real-World Contexts and Tasks will be presented this week at ACM IUI 2025. In it, we introduce the concept of Contextually-Driven Prompts (CDPs) which extend trigger-action programming to work with language-based AI models. We also present a system we built to test the idea, and report novel findings on how users initiate and refine policies for real-world tasks.

This project was led by my former intern Hai Dang at Reality Labs Research.

Check out the video below or see the full research paper.